MRI features of pediatric atypical teratoid rhabdoid tumors and medulloblastomas of the posterior fossa

Atypical teratoid rhabdoid tumor (AT/RT) occurs at a younger age and is associated with a worse prognosis than medulloblastoma; however, these two tumor entities are mostly indistinguishable on neuroimaging. The aim of our study was to differentiate AT/RT and medulloblastoma based on their clinical and MRI features to enhance treatment planning and outcome prediction.


| INTRODUCTION
Embryonal tumors account for approximately 20% of pediatric brain tumors, and they share a common histological feature: dense small round blue cells. 1,2 According to the 2021 World Health Organization (WHO) CNS5 classification, embryonal tumors can be further classified into "medulloblastoma" and "other CNS embryonal tumors". 1 Medulloblastoma is the most common posterior fossa brain tumor in pediatric patients, accounting for approximately 20% of pediatric central nervous system (CNS) tumors and 60% of embryonal tumors. 2,3 In comparison, atypical teratoid rhabdoid tumor (AT/RT), the leading subtype in the "other CNS embryonal tumor" category, accounts for 1%-2% of pediatric CNS tumors. 1,4 AT/RT develops at a younger age than medulloblastoma. Approximately 80.5% of AT/RTs occur in children under 3 years of age. 5 Due to the aggressive behavior of embryonal tumors, AT/RT and medulloblastoma are often treated with trimodality therapy consisting of surgery, chemotherapy and postoperative radiotherapy. 6 Radiotherapy is generally performed in patients older than 3 years of age due to associated neurocognitive toxicity in infants. [6][7][8] The prognosis of AT/RT is extremely poor, even though with a slight improvement in recent years. The 4-year overall survival rate was 43% with AT/RT (the ACNS0333 trial), compared to a 5-year overall survival rate of 82.3% in those with medulloblastoma (the SJMB03 trial). 9,10 AT/RT and medulloblastoma have many similar features and are generally indistinguishable on neuroimaging. 5,[11][12][13][14][15] More than half of the AT/RTs are located infratentorially. 14,15 On unenhanced computed tomography (CT), both tumors appear hyperdense due to the high nuclear-cytoplasmic ratio. 5,14,15 On magnetic resonance imaging (MRI), they appear iso-to-hypointense compared to the gray matter on T1-weighted imaging (T1WI), heterogeneous with variable signal intensity on T2-weighted imaging (T2WI) and show enhancement after contrast administration. 5,16 Intralesional hemorrhage may occur in both tumor types, although the incidence is higher for AT/RT. 5,14,15 Diffusion-weighted imaging (DWI) has been widely used for the differentiation of brain tumors 12,17 ; however, both AT/RT and medulloblastoma appear hyperintense on DWI and are reported to have an overlapping apparent coefficient (ADC) value. 12,14,18 Pretreatment diagnosis is important in treatment planning due to the significantly worse prognosis of AT/RT than medulloblastoma, especially in very young patients, for whom radiotherapy is less suitable. The aim of this retrospective study was to define clinical and MRI features that may help to differentiate AT/RT and medulloblastoma in pediatric patients to allow appropriate treatment planning and improve patient outcomes.

| METHODS
This retrospective study was approved and deemed exempt from individual patient consent for this research project by the institutional review board of our institute. Informed consent to perform imaging examinations, surgery and adjuvant cancer treatment was obtained from each patient or their family.

| Patients
From 2005 to 2021, there were 22 pediatric patients with histopathological diagnoses of AT/RT and 70 pediatric patients with medulloblastoma of posterior fossa in our institute. Six AT/RT patients and 13 medulloblastoma patients without presurgical MRI information in our hospital were excluded. In total, 73 pediatric patients (16 AT/RT and 57 medulloblastoma patients) were retrospectively enrolled in this study. We thoroughly reviewed their clinical characteristics, surgical records (gross total removal, nearly total removal, subtotal removal, or partial removal), adjuvant cancer treatments (chemotherapy and radiotherapy), tumor recurrence/progression, and survival.

| Presurgical MRI
Conventional MRI was performed with a 1.5T clinical MR scanner (Siemens Medical Solutions; GE Medical Systems; or Philips Medical Systems); the protocol included axial noncontrast T1/T2-weighted imaging (T1WI/T2WI), axial and sagittal contrast-enhanced RT from medulloblastoma. These distinct MRI findings together with the younger age of AT/RT patients may explain the worse outcomes in AT/RT patients.

K E Y W O R D S
atypical teratoid rhabdoid tumor (AT/RT), embryonal brain tumor, magnetic resonance imaging (MRI), medulloblastoma, pediatric brain tumor fat-suppressed T1WI, DWI with b values of 0 and 1000 s/ mm 2 applied in three orthogonal directions, and apparent diffusion coefficient (ADC) maps generated automatically by the MRI scanners. T2* gradient-echo (GRE) or susceptibility-weighted imaging (SWI for the Siemens instrument, SWAN for the GE instrument) was additionally performed in 15 patients. HWW and FCC reviewed all the images separately. They were blind to the clinical information and pathological diagnosis during the imaging analysis.
We analyzed the posterior fossa tumors according to their lesion size and signal intensity of the solid portions on T1WI, T2WI, and contrast-enhanced T1WI. Absolute ADC values (ADC min ) were measured by manually positioning regions of interest (ROIs) 10-50 mm 2 in size using hospital picture archiving and communication system (PACS) workstations. Tumor ROIs were positioned on the homologous area with the lowest signals within the solid components while avoiding areas with necrosis, peritumoral edema, calcification and hemorrhage. 12,13,17 If the tumor appears in more than three images, three ROIs were placed on different sections then averaged. If the tumor appears in less than three images, a total of three ROIs were placed within the tumor in the avoidance of overlapping. 13,14 To offset the subtle signal settings of different MRI scanners, an additional region of interest (ROI) was positioned on the homologous area of normal-appearing contralateral white matter of the cerebellum. 12,13,17 The ADC ratio was calculated as the solid tumor to contralateral white matter ratio. The DWI ratios (b = 1000 s/mm 2 ) were obtained in a similar manner but by positioning the tumor ROIs at areas with the highest signals ( Figure 1). 12,17 We further evaluated intratumoral morphology, including areas with necrosis, cysts, hemorrhage, and calcification as well as the dominant drainage veins for both tumor types. Intratumoral hemorrhage and calcification were recorded by comprehensively evaluating the T1WI, T2WI, DWI (b = 0 s/mm 2 ), GRE, and SWI findings. [19][20][21][22][23] Intratumoral hemorrhage was defined as signal alterations characteristic of hematoma and the presence of fluid-fluid levels ( Figure 2). 19 Most calcifications show low signals on both T1WI and T2WI, but some calcifications with large surface areas may appear hyperintense on T1WI. 20 In addition, both hemorrhage and calcification appear hypointense on DWI (b = 0 s/mm 2 ), GRE and SWI. [21][22][23][24] Suspected hemorrhage and calcification on MRI were further confirmed by CT, surgical reports and pathological findings.
The distribution of the main tumor drainage veins at either the central or peripheral location was evaluated on contrast-enhanced T1WI and T2WI (appearing as flow voids). The "tumor central vein sign" was defined as a single, dominant central intratumoral drainage vein that was clearly visible on contrast-enhanced T1WI and/or T2WI. The vein should be located centrally in the tumor, regardless of the lesion's shape, and may either appear as a thin line or dot. (Figures 3 and 4).
We also evaluated peritumoral involvement in pediatric patients with posterior fossa tumors. Peritumoral invasion into the adjacent brainstem and middle cerebellar peduncle was recorded by closely observing the fuzzy margins between the tumor and brain parenchyma on T2WI and contrast-enhanced T1WI. 25 (Figures 2 and  3) Peritumoral brain edema at the cerebellar peduncle or cerebellum was recorded. We also observed downward transforaminal extension of the tumor below the foramen magnum (below the McRae line), hydrocephalus (Evans' index >0.3) and leptomeningeal seeding ( Figure 3F). 24,26 Those with intraventricular drainage tubes were excluded from the analysis of ventricular dilatation.

| Postsurgical imaging studies
Postsurgical MRI findings were evaluated in all patients. The follow-up tumor status (recurrence/progression, stationary, or regression) was analyzed; two medulloblastoma patients with irradiation-induced gliomas were classified as having tumor progression. In those with tumor recurrence or progression, we evaluated the lesion patterns (local recurrence/progression, leptomeningeal seeding, or both).

| Statistical assessment
All statistical analyses were performed with IBM® SPSS® software. Continuous variables are summarized as the mean values with standard deviations; P values were calculated with the Mann-Whitney U test. Categorical variables are summarized as counts and percentages. P values were calculated with Pearson's chi-square or Fisher's exact test for factors with two categorical variables and with likelihood ratio tests for variables with more than 3 categories. Patient survival time was presented by using Kaplan-Meier method and P values were calculated with log-rank test. The optimal cutoff levels of the DWI ratio and ADC ratio for differentiating AT/RT and medulloblastoma were analyzed by receiver operating characteristic (ROC) curves and Youden's index. Patients with missing data for a variable were excluded from the analysis of that specific variable. All reported P values are two-sided. P values of less than 0.05 are regarded as statistically significant (in bold).

| Demographic features
The characteristics of the 16 AT/RT patients and 57 medulloblastoma patients are provided in Table 1. The two groups did not differ in terms of sex, surgery, or adjuvant treatment of the primary tumor. AT/RTs were diagnosed at a significantly younger age than medulloblastomas (2.8 ± 4.9 [0-17] vs. 6.5 ± 4.0 [0-18], p < 0.001).

| Clinical outcomes
During the follow-up period, AT/RTs had a significantly higher incidence of disease progression than medulloblastomas (13/16 [81.3%] vs. 20/57 [35.1%], p = 0.003). Among those with progressive disease, the time interval from surgery to tumor recurrence/progression was significantly shorter in patients with AT/RT (8.5 ± 10.6  months) than in patients with medulloblastoma (32.0 ± 40.9  months) (p = 0.004). No significant difference was observed in the recurrence pattern (local recurrence/ progression or leptomeningeal seeding) between the two groups (p = 0.86).
Two AT/RT and four medulloblastoma patients were lost to long-term follow-up. The overall survival time in the remaining patients was 45.5 ± 17.7  months in AT/RT patients and 128.4 ± 10.8 (10-187) months in medulloblastoma patients (p < 0.001). By the end of the study, AT/RT was associated with a significantly higher mortality rate than medulloblastoma (78.6% [11/14]

| MRI findings
MRI features of the 16 AT/RTs and the 57 medulloblastomas are provided in Table 2. In the presurgical MRI examination, no difference was observed in lesion size or signal intensity on T1WI, T2WI or contrast-enhanced T1WI between the two groups. DWI was performed in 15 AT/RT and 53 medulloblastoma patients. The raw data for calculating the ADC were accessible for 15 AT/    (Figures 1 and 5). In the ROC curve analysis, the ADC min cutoff value for differentiating AT/RT from medulloblastoma was 544.7 × 10 −6 mm 2 /s, with values under the cutoff indicative of AT/RT; the cutoff had 93.3% sensitivity and 67.3% specificity, and the area under the curve (AUC) was 0.842 ( Figure 5A,D). The ADC ratio had a cutoff value of 0.705 for differentiating AT/RT from medulloblastoma, with values under the cutoff indicative of AT/ RT; the cutoff had 86.7% sensitivity and 75.0% specificity, and the AUC was 0.857 ( Figure 5B,E). The DWI ratio had a cutoff value of 1.595 for differentiating AT/RT from medulloblastoma, with values under the cutoff indicative of AT/RT; the cutoff had 86.7% sensitivity and 64.2% specificity, and the AUC was 0.804 ( Figure 5C,F). No difference was seen in the ADC and DWI values of normal white matter between the groups.

| DISCUSSION
Although AT/RT and medulloblastoma of the posterior fossa have been reported to have similar clinical features and are nearly indistinguishable on neuroimaging, this study presents distinct clinical and MRI features that differentiate these two pediatric disease entities. 5,[11][12][13]27 In comparison to medulloblastoma, AT/RT has a higher progression/recurrence rate as well as a higher mortality rate. On MRI, AT/RT has a significantly lower ADC min and ADC ratio and a higher DWI ratio and is more prone to peritumoral invasion on T2WI. On the other hand, the "tumor central vein sign" had a significantly higher incidence in medulloblastoma than in AT/RT. DWI and the ADC are basic factors that are routinely obtained in MRI of brain tumors. By measuring the random movement of intralesional water molecules, their signal characteristics can reflect the cellularity and grading of various tumors. 12,17 Reported ADC values in young patients with AT/RT or medulloblastoma from an English literature review are listed in Table 3. [11][12][13][14]18,[28][29][30][31] Although the difference between embryonal tumors and other tumor types was disclosed, none of the studies were able to distinguish AT/RT from medulloblastoma using the ADC value, which may be related to the small number of cases. Gauvain et al, 28 Rumboldt et al, 13 Ahmeda et al 18 and Phuttharak et al 12 reported a lower ADC min or ADC ratio in AT/RT cases than medulloblastoma cases, which is consistent with our study. Notably, we are the first to demonstrate the differences statistically. The ADC min values reported by Koral et al 14 and Yamashita et al 29 were higher in AT/RT cases than in medulloblastoma cases, although only six and one AT/RT patients were included in their studies, respectively.
All AT/RT cases reported in the literature had an ADC min ranging from 0.45 to 0.85 (×10 −3 mm 2 /s) and an ADC ratio ranging from 0.63 to 0.87; our study revealed an ADC min of 0.47 ± 0.07 (0.35-0.58) and an ADC ratio of. 0.61 ± 0.11 (0.40-0.79) in AT/RT cases. The reported medulloblastoma cases had an ADC min ranging from 0.27 to 1.25 (×10 −3 mm 2 /s) and an ADC ratio ranging from 0.64 to 1.45; in comparison, our study revealed an ADC min of 0.57 ± 0.08 (0.33-0.80) (×10 −3 mm 2 /s) and ADC ratio of 0.77 ± 0.10 (0.52-0.95) in medulloblastoma cases. The ADC values measured in our study are quite similar but slightly lower than those in the literature. Notably, as we placed the tumor ROI at the most homogenous hypointense solid part of the lesion (Figure 1), the area of the ROI may be as small as 10 mm 2 . The benefit of our small ROI in measuring the most homogeneous hypointense part of the tumor includes its easy visualization and its representation of the most malignant tumor focus. Analyzing the most malignant tumor focus allows more accurate clinical tumor grading and diagnosis. In comparison, the areas of tumor Two patients with atypical teratoid rhabdoid tumors and four patients with medulloblastoma were lost to long-term follow-up and thus were excluded from the calculations.
ROIs in the literature were larger than that in our study and ranged from 30 to 100 mm 2 (Table 3). Large area measurements may include more uneven cell structures that result in a higher mean ADC min value. Furthermore, the normal reference used for calculating the ADC ratio differed between studies ( Table 3). The majority of the studies used cerebellar white matter as the reference, similar to our study. Additionally, many of our referential ROIs were positioned at the middle cerebellar peduncle (Figure 1). In contrast, some studies chose "cerebellar parenchyma" or "homogenous brain regions" as normal references instead of precisely placing the ROI at the white matter, which may correspondingly lead to a higher ADC ratio. The DWI ratio has been previously used by Wu et al 17 in tumor grading of brain gliomas, but it has never been applied in embryonal brain tumors. Phuttharak  created a "five-point visual scale" for analysis of the DWI intensity of posterior fossa tumors. All (4/4) AT/RTs were markedly hyperintense, while 25% (6/24) of medulloblastomas were hyperintense and 75% (18/24) of medulloblastomas were markedly hyperintense. This is consistent with our finding that AT/RTs were more hyperintense on DWI than medulloblastomas (DWI ratio: 1.77 ± 0.20 [1.47-2.26] vs. 1.52 ± 0.21 [1.15-1.96], p < 0.001). Both the DWI and ADC data of the present study support the notion that the tumor cell grade of AT/RT is higher than that of medulloblastoma.
A recent report by Zhang et al 27 successfully distinguished AT/RT from medulloblastoma via machine learning by analyzing MRI-based radiomic phenotypes, including their morphology and signal intensities on T2WI and contrast-enhanced T1WI. Their finding is key breakthrough, but these radiometric features are indiscernible to the human eye, and the high technical requirement limits routine clinical application. Therefore, we propose the following differentiating characteristics: the "tumor central vein sign" for medulloblastoma, a lower ADC value/ratio for AT/RT, and more aggressive marginal invasion for AT/RT. Our findings can be easily applied and interpreted in daily clinical practice and can assist future radiomics research.
Additionally, we found a higher incidence of intralesional hemorrhage in AT/RTs than in medulloblastomas, which is in accordance with a previous report by Koral et al. 14 In addition, AT/RTs were reported to have a higher cerebral blood volume (CBV) value than medulloblastomas on perfusion MRI by Goo et al, 32 indicating more abundant neovascularity in AT/RTs, which may support our finding. Abundant tumor vascularity may be unfavorable for complete surgical resection, thus influencing the outcomes of AT/RT patients, especially those who are very young.
A dominant central drainage vein was observed in 42.1% of the medulloblastomas but in only 1 AT/RT (6.3%). Although large draining veins in medulloblastomas were observed by neurosurgeons during tumor resection in one report, 33   diagnosis and avoid potential massive intraoperative tumor hemorrhage. In our study, AT/RT was associated with younger age, more aggressive clinical behavior and a worse outcome than medulloblastoma, these were compatible with the findings in previously reported literatures. 9,10 The dismal prognosis of AT/RT can be explained by several aspects observed in the present study: (1) The majority of AT/RTs occurred in very young children. Their unfused skull sutures accommodate the progressively increasing cranial pressure during tumor growth, which results in delayed symptom onset. 34 Even when symptoms are present, some nonspecific manifestations (irritability, vomiting, etc.) can be difficult to express due to the age of these very young patients, which results in diagnostic delay 34,35 ; (2) Surgery, chemotherapy and radiotherapy are regarded as the standard treatments for AT/RT. Maximal surgical resection and aggressive adjuvant therapy are particularly crucial in AT/ RT treatment due to its invasive behavior. 27 However, aggressive surgery followed by irradiation is sometimes postponed in patients under 3 years of age due to neurological side effects, which may lead to suboptimal treatment and a poor outcome. 36,37 (3) Significantly greater peritumoral invasion into the brainstem and cerebellum by AT/RT than by medulloblastoma was observed on T2WI, and AT/ RT can leave subclinical residual tumors after surgical resection. Without aggressive adjuvant treatment, these subclinical residual tumors at the tumor bed or brain surface may become the origin of tumor recurrence or leptomeningeal seeding. In addition, more tumor bleeding in AT/ RTs, as noted in the present study, and the reported higher tumor vascularity of AT/RTs may increase the technical difficulty of complete resection. (4) In the present study, on MRI, AT/RTs had significantly higher DWI ratios than medulloblastomas. A higher DWI value in CNS gliomas was proven to be associated with a higher tumor grade and a worse prognosis. 17,38 These aggressive tumor behaviors may explain the significantly lower survival rate in AT/RT patients than in medulloblastoma patients.
There are some limitations to our study. First, this was a retrospective study, with a relatively small number of AT/RT patients due to the rarity of the disease. Second, while statistically significant, the sensitivity and specificity of some radiologic findings may not reach high clinical significance. Surgical intervention and pathological diagnosis are still important for patients with AT/RT and medulloblastoma. Third, molecular analysis was not performed in our study. Of note, WNT/SHH subgrouping was not widely used until the pronouncement of the 2016 WHO classification of CNS tumors. 39 Further study is recommended. Forth, the MR images were obtained from different scanners, which may influence ADC and DWI measurements. Therefore, we further calculated the ADC ratios and DWI ratios to offset the different settings between scanners. Fifth, we did not evaluate more advanced MRI techniques, such as perfusion imaging and MR spectroscopy. 32

| CONCLUSION
AT/RT occurs at a younger age, has more aggressive clinical behavior, and has a more indistinct tumor margin than medulloblastoma, which results in a much worse prognosis. Maximal surgical resection and aggressive adjuvant therapy are therefore important in AT/RT treatment. We can differentiate AT/RT from medulloblastoma on MRI by the lower ADC value, higher degree of peritumoral invasion, and absence of the "tumor central vein sign," These MRI characteristics may be helpful in making pretreatment diagnosis and appropriate outcome evaluation.